Riccati conditioning and sensitivity for a MinMax controlled cable-mass system

Abstract

In this paper we present a numerical study that investigates the relationship between the parameter q, used in the design of the MinMax controller, and the conditioning of the approximate algebraic Riccati equations, the sensitivity of the eigenvalues of I-¿2P¿ to ¿ as well as the effect of q on the stability radia and the stability margin of the system. In order to guarantee accurate numerical solutions to the approximate Riccati equations, the Riccati equations must remain well-conditioned for the values of ¿ that are considered. This condition number reflects the combined sensitivity of the Riccati equations to the system inputs A, B, R, C and ¿. In addition, we also consider the sensitivity of the eigenvalues of I-¿2P¿ to ¿. We study the possibility of these sensitivities serving as an indication of the largest value of ¿ for which I-¿2P¿ remains positive definite. This sensitivity could also serve as an indication of the accuracy of the computation of I-¿2P¿. Lastly, in order to design efficient low order controllers, it is important to ensure the robustness of the design. Stability radius and stability margin serve as measures of the robustness of the controller. A one-dimensional nonlinear cable mass system is considered to illustrate these ideas and numerical results are presented.